Year: 2006
Type: article
Abstract: This paper presents a new connectionist model for serial order encoding and retrieval.It is based on a divergent-reconvergent structure, which encodes a list of items by superimposed distributed representations of each of the items, where early items are coded by more active units than late ones due to lateral inhibition.Thus, the first item is the strongest and the last the weakest.Retrieval is based on the activation gradient: selecting the most active item, and then inhibiting it to allow the next item to be retrieved.Side effects of this mechanism gives rise to the primacy effect, the recency effect and basic similarity effects, similar to those found in human immediate serial recall.The divergentreconvergent structure strikingly resembles the anatomical structure of the basal ganglia.This model provides a plausible computational and functional account for the sequencing function of the basal ganglia, and other brain areas that may implement it.
Authors: Elman, Jeffrey L., Shieh, Danke X
Institution: University of California San Diego
Language: English
Subfield: Cognitive Neuroscience
Field: Neuroscience
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W2625353413